A Rationale-based Model for Architecture Design Reasoning

3.3. Existing methods for capturing and representing design rationale
narrow, it saves screen real-estate but is difficult to hit with a mouse. Since the rationale
is based on argument and not proof, MacLean et al. argue that the elements in QOC
should be used to justify a design even though they may be subject to further arguments
[97]. As such, QOC can be expanded into any arbitrary level of elaboration. Designers
using QOC should select areas where these arguments serve the purpose of contentious
design but not expand into every possible detail because it is not useful to do so.
Figure 3.4: An Example of QOC Design Rationale
QOC supports the development of a space of alternatives. It enables designers to
consider criteria which could dictate viability of options. Unlike IBIS, PHI and REMAP
which capture the history of design deliberation, QOC focuses on design options.
Dutoit and Peach proposed the Rationale-based Use Case Specification method which
combines use case specification to the QOC method of argumentation [34]. The argumentation
for the use cases which comprised of functional and non-functional requirements
are then captured in an enhanced QOC model. This method provides a better integration
between requirement specification, design specification and the design rationale.
3.3.3 Design Rationale Language (DRL)
DRL records design rationale by describing how an artefact serves or satisfies expected
functionalities. DRL is an expressive language which represents the qualitative elements
in the reasoning spaces around decisions [94]. In DRL, the possible design options are
contained in the alternative space, and the arguments to support or contradict a design
is contained in the argument space, each design possibility is evaluated and the results
are contained in the evaluation space, the evaluation is performed according to certain
criteria which is contained in the criteria space, the issues which are made explicit and
containing the alternatives, evaluations and criteria are contained in the issue space. The
fundamental object types in DRL are goal, question, claim and alternative. The structure
of a decision graph using these elements is shown in Figure 3.5.
A goal represents the criteria that needs to be satisfied. An alternative represents
33